rx4pain,
RE AMSC stochastics:
To me, the 13-7-7 stochastic is the better indicator of the three.
Lets look a bit closer. First, the 15-5-5:
askresearch.com
In the 3 month chart, there are a total of 8 crossovers (4 buy signals, 4 sell signals). The first buy signal was a good one, but the subsequent sell signal came late. The second buy signal was also a good one, but the sell signal again came late. The third buy signal came late, but the subsequent sell signal got you out quickly, and so was a good one. The fourth buy signal was a good one, but again the sell signal came late, after that long black candle. So, out of 8 buy and sell signals, we have 4 out of 8 good ones (50%). There were 3 out of 4 good buy signals (75%), and only 1 out of 4 good sell signals (25%).
Now lets do the same thing with the 5-3-3 stochastic setting:
askresearch.com
This time there were a total of 15 signals (7 buy, 8 sell). Going from left to right, I would classify these as good, bad, good, good, bad, good, good, bad, bad, bad, good, good, good, and bad. Total, 8 good, 7 bad (53%); 5 out of 8 (62%) sell signals were good, and 3 out of 7 (43%) buy signals were good.
Now, let's look at the 13-7-7:
There are a total of 6 signals, 3 buy signals and 3 sell signals. I would classify these as good, bad, good, good, good. The fourth signal on 12/18 came late, but persisted, keeping you out of the long position during a short-lived minor rally the last week of December/first week of January, but getting you back in very nicely just after the bottom. So, I would classify these as 3 out of 3 good buy signals (100%), and 2 out of 3 good sell signals (66%). Total, 5 out of 6 signals were good (83%).
Summarizing, my assessment is that if one were to pick just one stochastic setting to use, the ones which would give the best overall probability of good buy and sell signals would be
13-7-7: 83%
15-5-5: 50%
5-3-3: 43%
So, the 13-7-7 seems superior to me overall, with the 15-5-5 superficially similar to the 5-3-3. However, the 5-3-3 gives way more signals, so there were more chances to lose, just to pick up rather small gains.
Here's how each performed with buy signals:
13-7-7: 100%
15-5-5: 75%
5-3-3: 43%
and with sell signals:
13-7-7: 83%
5-3-3: 62%
15-5-5: 25%
Probably the optimal thing here would be to use the 13-7-7 as a buy signal (which would get you in reasonably early, and without false signals), and then use the 5-3-3 as a sell signal.
Looking at the chart, that would produce transactions which I guestimate would have gone like this (roughly):
11/24 Buy at 30; 11/28 sell at 27; P/L -3, or -10%
11/30 Buy at 26; 12/12 sell at 34; P/L +8, or +31%
1/9 Buy at 24; 1/19 sell at 32; P/L +8, or +33%
Total gain +13 points, or +43% compared to the starting price of 30. During this time, AMSC began at 30 and ended at 32, for a net change of +2, or 7%.
I realize there is lots to quibble over here, but I think the point is valid: with this stock, the 13-7-7 stochastic gives the best overall accuracy of signals, and the 5-3-3 stochastic probably does the best job of getting you out with minimal losses, but results in significant time out of the stock when used in conjunction with the 13-7-7 as a buy signal.
From a practical standpoint, once you have decided which stochastic setting gives the best signals, then the approach which will tend to give the best overall results with time will usually be to use that signal to get you in (in this case, the 13-7-7), and a faster responding signal to get you out at the first sign of trouble (i.e, the 5-3-3). This will tend to result in a higher batting average, but smaller gains per trade, and probably significant time out of the stock. But the advantage is that few trades will be tend to be losers.
Now, the 13-7-7 has not signaled buy yet, and appears to be several days away from a signal.
JMVHO......YMMV
WS
P.S. I would encourage you and everybody else to paper trade these approaches going forward, and report what you find. In fact, I'll try to do just that with this particular stock, AMSC. |
